Titlacahuan
this is going to be hours of fun: github-roast.pages.dev
Titlacahuan
sorry eyedeekay, had to roast you:
Titlacahuan
Eyedeekay, or should I say, master of the meaningless hustle? You've got more repos than a bot on a sugar rush, yet your stargazer count is lower than your mom’s WiFi signal. Seriously, you’re into "making trivial activities enjoyable over anonymous networks"? Well, congratulations on being the king of creating pointless apps that nobody ever uses. Half your repos are "forked"—so basically, you
Titlacahuan
're a glorified copy-paster. Next time, try creating something original. Then again, with a handle like "idk", what more should anyone expect? Stay anonymous, champ—it's the only thing you’re good at.
dr|z3d
Titlacahuan: you've been tolerated here for a while and seemingly kept your end of the bargain by not insulting or threatening anyone. To say you're welcome here would be a gross overstatement. Ad hoc insults directed at anyone, not least eyedeekay, cannot, will not, be tolerated. Sorry.
orignal
zzz, I have got some new info
orignal
an adversary is trying to recognize i2p traffic by number of new TCP/UDP connections per interval
orignal
I thing we should add random variance to session timeout
zzz
please explain more what the threat model is
orignal
they measure number of new connections(probably TCP only) say per 10 minutes
orignal
and see that's more of less unique for TCP vs. say torrents
zzz
that seems like a pretty difficult way to do protocol identification. Have you seen some research about it?
orignal
the ratinaly of this that we have constant session tiemout
orignal
no, it's just rumor
orignal
but do you think it doesn make sense if you see a p2p protocol measure connection pattern?
zzz
in java we have a constant, but it's "sloppy" because we don't have a per-session timer, we have a single expiration job that runs periodically
zzz
sure, p2p looks a lot different than client/server, but there's lots of p2p protocols
orignal
say I see TCP connection trminatiing excatly after 120 seconds of inacctivity
orignal
I would suppose it's i2p
zzz
but it's not exact. see above.
orignal
and if you see majority of TCP session terminting around this value
zzz
looking at packet sizes would be a lot easier
orignal
not exact but with relatively small deviation
orignal
same in i2pd
zzz
or just look at if it's both TCP and UDP. how many p2p protocols do both? probably not too many
orignal
yes that;'s another possibility
orignal
but constant termination timeout is too obvious
zzz
your hypothetical network observer is very powerful to track packet interarrival times for a zillion IP pairs
orignal
say they sit at ISP
orignal
btw, in Russia they detect wiregaurd this way
orignal
by measuring handshake intervals
zzz
interesting. handshake timing/size sounds a lot easier than termination/timeout timing. have a link to any research?
orignal
about wg?
zzz
yes, detecting wireguard with handshake intervals
cumlord
original did you make updates to i2cp?
cumlord
i'm seeing 1MB/s on i2pd right now
orignal
somewhere at habr
orignal
yes I keep improving it
orignal
that's the link
zzz
thx
cumlord
btw, this is the special character thing i'm seeing in snark, torrents made with mktorrent. think the issue could be from metata in .torrent file?
cumlord
seems a lot better original, i2pd seems like it's keeping up with my i2p+ routers
zzz
whats the issue?
cumlord
in snark+ standalone, giving it a completed multi file torrent with .torrent file created with mktorrent
zzz
oh and it can't find the Gra*ce files?
zzz
Gr^ace
cumlord
filename in question is "The 100 - S02E11 - Coup de Grâce Bluray 1080p HEVC DDP 5.1-SPANK.mkv" but i've seen it happen a couple other times with special chars. It doesn't recognize the filename and i think makes a socks file of the same name
cumlord
yes
zzz
win or lin?
cumlord
lin
zzz
UTF-8 locale presumably?
cumlord
different behavior in embedded snark, before i've been able to get it to find by replacing with a "_"
cumlord
yes
zzz
so it works in embedded?
zzz
what is a "socks file"?
cumlord
Actually no, it doesn’t work in embedded it just replaces with a “?”
cumlord
Weird that was the workaround Ive used before
cumlord
Sorry socket file
zzz
plus or vanilla?
zzz
we don't do socket files in java
zzz
so if you go to the torrent dir do you see Gr^ace files or Gr_ce files or both, or what?
cumlord
plus
cumlord
ah yes that’s not what that is, I see 2 files with the right chars, same name
cumlord
Both as “ The 100 - S02E11 - Coup de Grâce Bluray 1080p HEVC DDP 5.1-SPANK.mkv”
zzz
oh yeah duh that's in the screenshot
zzz
so probably one has the ^a encoded differently from the other
zzz
try something like ls *S02E11* | hd
zzz
you started with mktorrent to create the .torrent file and passed it to snark?
cumlord
ah that might be it i'll add the original file back and check that
zzz
well there's a couple ways to fix it, but if you pick the wrong one, the leechers will have problems
zzz
maybe
zzz
if the original file name is not UTF-8 that's the issue
zzz
so do ls *S02E11* | hd and paste here
zzz
and then we'll figure out what's different and which one is which
cumlord
yes all my torrents are made with mktorrent
cumlord
the file snark likes shows up with "Gra..ce" with ls | hd
zzz
paste here for both files and let's see
zzz
need the hex
cumlord
ah, if you need me to paste that in irc i can, client turns into pastebin if too many lines
zzz
i got it
zzz
so the first is c3 a2
zzz
the second is 61 cc 82
cumlord
ok i see what you're saying
cumlord
if that's modified think it would bork leechers?
zzz
c3 a2 is valid/correct UTF-8 en.wikipedia.org/wiki/%C3%82
zzz
cc 82 seems to end up as a ^ after the 61 (a) that's a unicode combining character
Quaddle
h
zzz
so they're BOTH valid UTF-8, but different
cumlord
Thanks for looking, that’s strange
zzz
so now, hd the torrent file and tell us which encoding it has in it
zzz
that will tell us who is to blame
cumlord
Sorry gtg for now be back to try that
zzz
ok, I'll play around with it
zzz
something is not surviving the round trip bytes->chars->bytes, the culprit is UTF-8 "normalization"
zzz
quick test on both flavors seems like snark is happy with both and doesn't do any byte-changing, but not promising it's bug-free
eyedeekay
zzz I'm going to squint at the implications of subDb's for this issue: git.idk.i2p/i2p-hackers/i2p.i2p/-/issues/487
eyedeekay
*subDbs